merge_transformed_page: Allow transforming annotation rectangles

[vitek]

I'm using pdfly to create a book from single-page pdf file. Link on the right side does not work.

I've nailed it down. When attributes are merged /Rect is not translated. I made a workaround in my own script:

...
        page = reader.pages[first]
        other = reader.pages[second]

        tx = float(page.mediabox.width)
        other[NameObject(PG.ANNOTS)] = ArrayObject(
            list(
                _transform_annots(get_annotations(other), tx=tx)
            )
        )

        page.merge_translated_page(
            other, tx=tx, ty=0, expand=True,
        )
        writer.add_page(page)

def _transform_annots(annots, tx):
    for annot in annots:
        obj = annot.get_object()
        #obj = copy.deepcopy(obj)
        if '/Rect' in obj:
            rect = obj['/Rect']
            obj[NameObject('/Rect')] = ArrayObject([
                FloatObject(rect[0] + tx),
                rect[1],
                FloatObject(rect[2] + tx),
                rect[3],
            ])
        yield annot


def get_annotations(page):
    if PG.ANNOTS in page:
        annots = page[PG.ANNOTS]
        if isinstance(annots, ArrayObject):
            return annots
    return []

I'd like to have this fixed in upstream version of pypdf. I tried to do it myself and ran into a problem. I can translate existing indirect object (as in code above). But it should be copied since attribute is mutated. And I'm not sure how to do it.

[stefan6419846]

Please provide a proper report including runnable example code and a corresponding PDF file. In general, this is similar to #2592, although more complex as the transformation matrix is not restricted to one operation, but could be everything - this is why your current approach would not be sufficient as well and most likely require proper rectangle transformations.

I tried to do it myself and ran into a problem. I can translate existing indirect object (as in code above). But it should be copied since attribute is mutated. And I'm not sure how to do it.

You probably want to replace obj[NameObject("/Rect")].get_object() in-place.

[vitek]

Please provide a proper report including runnable example code and a corresponding PDF file.

from pypdf import PdfReader, PdfWriter, PageObject
from pypdf.constants import PageAttributes as PG, AnnotationDictionaryAttributes
from pypdf.generic import NameObject, ArrayObject, FloatObject


def main():
    reader = PdfReader('test.pdf')
    writer = PdfWriter()

    page = reader.pages[0]
    other = reader.pages[1]
    tx = float(page.mediabox.width)

    page.merge_translated_page(
        other, tx=tx, ty=0, expand=True,
    )
    writer.add_page(page)

    with open('test-out.pdf', "wb") as fp:
        writer.write(fp)


main()

test.pdf

pdf file is attached. script creates test-out with 2-pages per sheet, left link is clickable while right isn't.

In general, this is similar to #2592

You're right. Sorry for creating issue duplicate.

although more complex as the transformation matrix is not restricted to one operation, but could be everything - this is why your current approach would not be sufficient as well and most likely require proper rectangle transformations.

Yeah, I'm aware of this. In my case I only need offset.

I tried to do it myself and ran into a problem. I can translate existing indirect object (as in code above). But it should be copied since attribute is mutated. And I'm not sure how to do it.

You probably want to replace obj[NameObject("/Rect")].get_object() in-place.